The non-Hermitian minor process
2026
Read The non-Hermitian minor process (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-21265.txt.
Introduction
Giorgio Cipolloni, László Erdős, and Oleksii Kolupaiev study log-determinants of leading principal minors of large non-Hermitian random matrices. They show these converge to a two-plus-one-dimensional Gaussian field. The field is logarithmically correlated for the parabolic distance, like the Edwards-Wilkinson class. The result links random matrix minors to a well-studied interface growth field.
Catalog record
| YEAR | 2026 |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 122013 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.21265
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The non-Hermitian minor process Abstract . We show that the log-determinant of leading principal minors of large non-Hermitian random matrices converges in distribution to a 2+1 dimensional Gaussian field, which is logarithmically correlated for the parabolic distance, reminiscent to the Edwards-Wilkinson universality class. Giorgio Cipolloni 1 1 1 University of Rome Tor Vergata, Via della Ricerca Scientifica 1, 00133 Rome, Italy. [email protected] László Erdős 2 2 2 Institute of Science and Technology Austria, Am Campus 1, 3400 Klosterneuburg, Austria. [email protected] Oleksii Kolupaiev 2 2 2 Institute of Science and Technology Austria, Am Campus 1, 3400 Klosterneuburg, Austria. [email protected] Keywords: Local Law, Zigzag Strategy, Dyson Brownian motion, Girko’s Formula, Linear Statistics. 2020 Mathematics Subject Classification: 60B20, 82C10. 1. Introduction We consider an N\times N matrix X with independent, identically distributed (i.i.d.) centered entries and its leading principal minors , i.e. square sub-matrices in the lower right corner 1 1 1 We deviate from the usual terminology where leading principal minors refer to minors in the top left corner. . We study the joint distribution of their eigenvalues through the correlation structure of their linear eigenvalue statistics (see ( 1.1 ) below), in particular, we describe how the different minor levels influence the answer. To motivate this line of study, we first summarize the analogous results for the Hermitian case, where there has been a lot of progress. Then, we will explain our result on minors of non-Hermitian random matrices. In particular, the main novelty of this work is to investigate this natural question in the non-Hermitian setting. Somewhat surprisingly, our result is new even for the Ginibre ensemble , i.e. for the Gaussian case, where otherwise many explicit formulas are available. Let now W be a Wigner matrix , i.e. an N\times N Hermitian random matrix with i.i.d. centered entries (up to the Hermitian symmetry) with \mathbf{E}|W_{ij}|^{2}=N^{-1} , and denote its eigenvalues by \lambda_{1},\dots,\lambda_{N} . It is well known since [ 43 ] that the (random) empirical measure \frac{1}{N}\sum_{i}\delta_{\lambda_{i}} converges to a limiting deterministic measure \mu_{\mathrm{sc}}(\mathrm{d}x) (i.e. the semicircular distribution ). It is then natural to ask for the size of the fluctuations around this deterministic limit. In [ 2 , 3 , 33 , 36 , 40 ] (see also [ 39 , 41 ] for later results) it was discovered 2 2 2 In fact, this unusually small fluctuations were discovered earlier by Dyson and Mehta [ 25 ] when studying the fluctuations of the eigenvalues of Haar unitary random matrices (CUE ensemble). that the fluctuations around \mu_{\mathrm{sc}}(\mathrm{d}x) are of size N^{-1} , i.e. unusually small. This is due to the very strong correlation among the eigenvalues of W , if the eigenvalues were independent or weakly correlated, the usual CLT would predict a much larger fluctuation of order N^{-1/2} . The fluctuation, however, is still Gaussian on this small scale. More precisely, we have L_{N}(f):=\sum_{i}f(\lambda_{i})-\mathbf{E}\sum_{i}f(\lambda_{i})\stackrel{{\scriptstyle N\to\infty}}{{\Longrightarrow}}\mathcal{N}(0,V_{f}), for any regular test function f , with an explicit V_{f}>0 . In particular, the limiting field can be identified as a one-dimensional logarithmically correlated (or log-correlated , in short) field. Instead of a single matrix W it is then natural to consider the family of its leading principal minors 3 3 3 Here W^{(k)} denotes the k\times k matrix obtained from W by removing its first N-k rows and columns. W^{(k)} for different k ’s to study the joint distribution of their eigenvalues, and show that these fluctuations converge to a two-dimensional Gaussian field. In [ 7 ] Borodin considered the joint spectral statistics of an N\times N Wigner matrix W=W^{(N)} and its minors W^{(xN)} , for x\in(0,1] , where we assume that xN is an integer for notational simplicity. In this work, Borodin showed that \sum_{i}f(\lambda_{i}^{x},x)-\mathbf{E}\sum_{i}f(\lambda_{i}^{x},x) , with \lambda_{i}^{x}\in\mathrm{Spec}(W^{(xN)}) and x\in(0,1] , converges to the two-dimensional Gaussian Free Field (GFF) , which can be realized by h_{N}(E,x)=\log|\mathrm{det}(W^{(xN)}-E-\mathrm{i}0)| in the sense that L_{N}(f)=\langle h_{N},f\rangle_{H^{1}} after a convenient change of variables. Since this fundamental work of Borodin there has been a great interest in joint fluctuations of analogous two-dimensional processes 4 4 4 We refer to the introduction of [ 9 ] for a detailed description of all the results connecting random matrices and logarithmically correlated fields. Here we focus only on the results concering minors or analogous processes defined for certain \beta -ensembles. , e.g. for sample covariance matrices [ 24 ] , \beta -Jacobi ensembles [ 5 ] , the adjacency matrix of random regular graphs [ 34 , 27 ] , and random tilings [ 8 ] and references therein. We also mention a related line of work studying the difference of linear eigenvalue statistics of a matrix and its immediate minor obtained by removing one row and column, which was initiated in [ 29 ] for the \beta -Jacobi ensemble and then, motivated by this work, studied for Wigner matrices [ 38 ] and for sample covariance matrices [ 23 ] . The outcome can be interpreted as convergence to the derivative of a GFF. Furthermore, [ 6 ] studied the joint correlations of eigenvalues of minors for a dynamical version of Wigner matrices giving rise to a three dimensional log-correlated field (with respect to the Euclidean distance) whose certain two dimensional slices are GFF. While there has been great interest and success in studying the joint eigenvalue fluctuations of minors of Hermitian random matrices, surprisingly, no results are known in the non-Hermitian case, not even for the Ginibre ensemble 5 5 5 We say that an N\times N matrix X belongs to the real/complex Ginibre ensemble if the entries \sqrt{N}X_{ij} are distributed as i.i.d. standard real/complex Gaussians. , i.e. when the matrix entries are independent Gaussian. While all our results in this paper hold for general non-Hermitian matrices with either real or complex i.i.d. entry distribution, for simplicity of this introduction, here we present only the complex Gaussian model. For the results in full generality, see Section 2 below. We thus define the Ginibre minor process , i.e. we consider an N\times N complex Ginibre matrix X and its leading principal minors 6 6 6 We again assume that xN is an integer, for notational simplicity. X^{(xN)} , for x\in(0,1] . In particular, for any smooth function f:\mathbf{C}\to\mathbf{C} , we study the joint fluctuations of the (centered) linear statistics (1.1) L^{(xN)}(f):=\sum_{i=1}^{xN}f\big(\sigma_{i}^{(xN)}\big)-\mathbf{E}\sum_{i=1}^{xN}f\big(\sigma_{i}^{(xN)}\big), where \sigma_{i}^{(xN)}\in\mathrm{Spec}(X^{(xN)}) , for different x\in(0,1] . (a) (b) Figure 1 . In the left panel we depict the eigenvalues of X^{(xN)} at several levels of the non-Hermitian minor process. By Girko’s circular law, almost all of these eigenvalues are contained inside of the parabolic surface |z|^{2}=x . The right panel illustrates the set-up of Theorem 1.1 . The dark grey disks chosen at the levels x and y with x\leq y , represent the supports of the test functions f and g , respectively. Theorem 1.1 . Consider the complex Ginibre minor process X^{(xN)} , and fix any x,y\in(0,1] with x\leq y . Let f,g be smooth functions supported in the open disks 7 7 7 Here \mathbf{D}_{x}:=\{z\in\mathbf{C}:|z|\leq\sqrt{x}\} . \mathbf{D}_{x} , \mathbf{D}_{y} , respectively. Then, the linear statistics L^{(xN)}(f),L^{(yN)}(g) converge jointly to centered Gaussian random variables L(x,f),L(y,g) , with covariance (1.2) \mathbf{E}L(x,f)L(y,g)=-\frac{1}{\pi^{2}}\int_{\mathbf{D}_{x}}\!\!\mathrm{d}^{2}z\int_{\mathbf{D}_{y}}\!\!\mathrm{d}^{2}w\,\,\overline{\partial_{z}f(z)}\partial_{w}g(w)\partial_{z}\partial_{\overline{w}}\log K\big((x,z),(y,w)\big), where K is given by 8 8 8 Later in ( 2.12